By continuing to use our site, you consent to the processing of cookies, user data (location information, type and version of the OS, the type and version of the browser, the type of device and the resolution of its screen, the source of where the user came from, from which site or for what advertisement, language OS and Browser, which pages are opened and to which buttons the user presses, ip-address) for the purpose of site functioning, retargeting and statistical surveys and reviews. If you do not want your data to be processed, please leave the site.

Replica Synthesis Software

Replica Synthesis software ingests real data and builds data synthesis models to generate high utility synthetic datasets.

Critical features of Replica Synthesis include:


  • Customizable synthesis parameters.
     
  •  It can be deployed on the cloud or on-premises.
     
  • Clients do not have to share their data to synthesize it.
     
  • Comprehensive REST API for integration with multiple and varied front ends.
     
  • It generates detailed data utility results.
     
  • Produces a synthesis report which describes the data, methodology, the synthesis results, the utility results, and any limitations. The report template can be easily customized.
     
  • The software is built for small and large datasets.
     
  • SDKs supporting multiple data science and software engineering end-users.
     
  • Flexible synthesis plan specifications to handle complex datasets.
     
  • Easy deployment through containers and virtual machine images.

How is Synthetic Data Generated?


The main approach for generating synthetic data is to have a generator model and a discriminator function.

The generator takes real datasets, and using various statistical machine learning and deep learning models, generates synthetic data.

The discriminator evaluates how good that synthetic data is compared to the real data. If the differences between them are large then this information is fed back and a new generator model is trained to try to narrow these differences. Therefore, it is an iterative process of training generator models until they produce acceptable synthetic data.

The highly automated Replica Analytics pipeline will produce the datasets as well as comprehensive utility assessments for the data. This helps with creating buy-in from data analysts and other data users. Privacy Assurance Services from Replica Analytics complement the main synthesis technology to provide the necessary evidence to the legal team.   

For more information on Replica Synthesis software, please email info@replica-analytics.com